Farms, Fabrics and Clouds Steve Loughran Julio Guijarro HP Laboratories, Bristol, UK December 2007 steve.loughran@hpl.hp.com julio.guijarro@hpl.hp.com
Julio Guijarro Researcher at HP Laboratories Area of interest: Deployment In charge of OSS release http://smartfrog.org/ Farms, Fabrics and Clouds www.smartfrog.org Page 2
Steve Loughran Researcher at HP Laboratories Area of interest: Deployment Author of Ant in Action Farms, Fabrics and Clouds www.smartfrog.org Page 3
Our research • How to host big applications across distributed resources – Automatically – Repeatably – Dynamically – Correctly – Securely • How to manage them from installation to removal • How to make dynamically allocated servers useful Farms, Fabrics and Clouds www.smartfrog.org Page 4
Question Who had breakfast this morning? Farms, Fabrics and Clouds www.smartfrog.org Page 5
Question Who harvested wheat or corn, or killed an animal for that breakfast? Farms, Fabrics and Clouds www.smartfrog.org Page 6
Farms provide food. It is somebody else's problem Farms, Fabrics and Clouds www.smartfrog.org Page 7
Question Who is wearing clothes they wove or knitted themselves? Farms, Fabrics and Clouds www.smartfrog.org Page 8
Provisioning of clothing -fabrics- is outsourced It is somebody else's problem Farms, Fabrics and Clouds www.smartfrog.org Page 9
All new applications are on the Web • Web Browser, AJAX clients • Richer: Flash, XUL, Silverlight • "… as a Service " ⇒ Lots of code running in the server ⇒ Data mining/analysis problems ⇒ Unpredictable demand Farms, Fabrics and Clouds www.smartfrog.org Page 10
Old world installation: single server ������������������ �������� Single web server, Single DB ��������� RAID filestore ������ -SPOF -limitations of scale Farms, Fabrics and Clouds www.smartfrog.org Page 11
yesterday: clustering ��������� ������ Multiple web servers, Replicated DB ������������ ������� RAID Network filestore Load-balancing router -Cost -Complexity ������������ �������� -Limitations of scale ��������� �������� Maintains the illusion of a single server Farms, Fabrics and Clouds www.smartfrog.org Page 12
Now: server farms 500 web servers, Distributed filestore Rented storage & CPU Scales up No capital outlay Agile infrastructure Farms, Fabrics and Clouds www.smartfrog.org Page 13
tomorrow? grid fabric. 50000 servers Farms, Fabrics and Clouds www.smartfrog.org Page 14
Application architectures and deployment problems change radically in this world Farms, Fabrics and Clouds www.smartfrog.org Page 15
Application architectures • ROA/REST • Virtualized • Map/Reduce • Shards • Tuple-spaces • Grid Farms, Fabrics and Clouds www.smartfrog.org Page 16
Virtualization Farms, Fabrics and Clouds www.smartfrog.org Page 17
Why? • Save on hardware (and power, space) • Dynamically move running servers • Demand creation of new images • Testing complex system configurations • Redistributing entire machine image • 'virtual appliance' Farms, Fabrics and Clouds www.smartfrog.org Page 18
Assumptions that are now invalid • Systems have a long lifespan • It is slow/expensive to create a new system • It is expensive to duplicate one • Systems can/should be managed by hand • Clocks proceed at the same rate • Physical RAM doesn’t get swapped out • Running machines can't be moved/cloned Farms, Fabrics and Clouds www.smartfrog.org Page 19
Server Farms Farms, Fabrics and Clouds www.smartfrog.org Page 20
Assumptions that are now invalid • System failure is an unusual event • 100% availability can be achieved • Data is always near the server • You need physical access to the severs • Databases are the best storage form • You need millions of $/£/€ to play Farms, Fabrics and Clouds www.smartfrog.org Page 21
Who has the servers? • Yahoo, Google, MSN, eBay: services • MMORPG Game Vendors: Word of Warcraft, Second Life • EU Grid: Scientists • HP, IBM, Sun: rent to companies -focus on CPU performance • Amazon: rent to anyone with an Amazon account -focus on startups Farms, Fabrics and Clouds www.smartfrog.org Page 22
Amazon EC2 Public Internet $ $ $ $ Host Host AMI AMI AMI AMI $ (Xen VM) (Xen VM) (Xen VM) (Xen VM) /mnt /mnt /mnt /mnt Fast (free) network free access; slow initial read time pay per GET; per megabyte S3 Storage Farms, Fabrics and Clouds www.smartfrog.org Page 23
Amazon EC2 • Pay as you go Virtual Machine Hosting • No persistent storage other than S3 filestore - uses HTTP GET/PUT/DELETE operations • $0.10 per CPU/hour • S3 Storage has own billing (by MB & by access -cheaper in bulk) Farms, Fabrics and Clouds www.smartfrog.org Page 24
Demo Farms, Fabrics and Clouds www.smartfrog.org Page 25
Map/Reduce Run code near the data, then merge the results Farms, Fabrics and Clouds www.smartfrog.org Page 26
Assumptions that are now invalid • Terabyte datasets are hard to work with • Code runs on a single machine • Sequential code is better than parallel code • RAID hardware is the best way to store data • Databases are better than filesystems Farms, Fabrics and Clouds www.smartfrog.org Page 27
Shards Farms, Fabrics and Clouds www.smartfrog.org Page 28
Assumptions that are now invalid • A single farm needs to scale to infinity • You need to provide 100% availability to 100% of users • You have to roll out simultaneous updates to the application, changes to the DB schema, globally Farms, Fabrics and Clouds www.smartfrog.org Page 29
Changes for developers • Many classic assumptions are invalid • Design for scale • Rent servers from the outset —every developer can have their own set • Cover your server costs from the outset and you are in the black from day 1 Farms, Fabrics and Clouds www.smartfrog.org Page 30
Problems for us farmers • Power management • Predictive disk failure management • Load balancing for availability, power • Data cache management • Billing • Security/Isolation • How this will change server hardware • Managing/Configuring Machine Images • Diagnostics when things go wrong Farms, Fabrics and Clouds www.smartfrog.org Page 31
Topic for discussion Where is all this heading? Farms, Fabrics and Clouds www.smartfrog.org Page 32
Recommend
More recommend
